.png){kind=icon}
.png){kind=icon}
Return
A consistency regularization training method for automatic modulation classification under incomplete information
.png){kind=icon}
PRE
AI
DOI：10.1016/j.patcog.2025.112655.png)
.png){kind=icon}
Original request for help.png){kind=icon}
Share.png){kind=icon}
Save
Abstract
En 中文
• A consistency-regularized training framework is proposed for robust automatic modulation classification (AMC) under incomplete signal patterns. The method introduces a novel consistency regularization loss to alleviate output distribution mismatch caused by Dropout, enhancing model generalization in scenarios with scarce or biased training data. • A lightweight one-dimensional residual feature extraction network is designed to balance computational efficiency and representational capacity. By leveraging shared 1D convolutions and residual connections, the model effectively captures discriminative modulation features from raw I/Q signals while maintaining low complexity. • Extensive experiments across three benchmark datasets (RadioML2016.10a, RadioML2016.04c, and RadioML2022) demonstrate that the proposed CRCNN outperforms state-of-the-art methods, especially under low-data and low-SNR conditions. The CRCNN model achieves significant gains in accuracy and robustness, showing strong generalization with as little as 1 % of training data.
Journal
IF:
7.6
Papers:
1.3W
Citations:
4.5W
View more organizations